How to interconnect for Massive MIMO Self-Calibration?

Fuqian Yang, Hanyu Zhu, Cong Shen, Linglong Dai, Xiliang Luo

Abstract

In time-division duplexing (TDD) systems, massive multiple-input multiple-output (MIMO) relies on the channel reciprocity to obtain the downlink (DL) channel state information (CSI) with the acquired uplink (UL) CSI at the base station (BS). However, the mismatches in the radio frequency (RF) analog circuits at different antennas at the BS break the end-to-end UL and DL channel reciprocity. To restore the channel reciprocity, it is necessary to calibrate all the antennas at the BS. This paper addresses the interconnection strategy for the internal self-calibration at the BS where different antennas are interconnected via hardware transmission lines. Specifically, the paper reveals the optimality of the star interconnection and the daisy chain interconnection respectively. From the results, we see the star interconnection is the optimal interconnection strategy when the BS are given the same number of measurements. On the other hand, the daisy chain interconnection outperforms the star interconnection when the same amount of time resources are consumed. Numerical results corroborate our theoretical analyses.

The Challenge of Channel Reciprocity

Massive MIMO systems are a cornerstone of modern wireless communication, leveraging large antenna arrays at the base station to improve spectral efficiency. In TDD mode, these systems assume that the physical propagation channel is identical for both uplink and downlink. This "channel reciprocity" allows the BS to estimate the downlink channel using uplink pilots, significantly reducing overhead.

In practice, the end-to-end channel includes the RF front-ends. Differences in the transmit (Tx) and receive (Rx) chains across different antennas introduce gain and phase mismatches. Internal self-calibration addresses this by interconnecting antennas via hardware transmission lines, allowing them to exchange signals and estimate these relative mismatches. The fundamental design question is: how should we wire these antennas together?

Interconnection Strategies: Star vs. Daisy Chain

The paper evaluates two primary wiring strategies: the Star Topology, where every antenna connects directly to a single central reference antenna, and the Daisy Chain Topology, where antennas are connected sequentially in a line.

Daisy Chain Interconnection Topology

The authors use the Cramer-Rao Lower Bound (CRLB) to analytically determine the best accuracy achievable by any estimator. Under the assumption of identical transmission lines and equal RF gain amplitudes, they found that the star topology is measurement-optimal. Because every antenna is only one "hop" from the reference, estimation errors do not accumulate as they might across a long chain.

However, the star topology suffers from a significant temporal bottleneck: measurements must be performed sequentially to avoid interference, leading to a total calibration time that scales linearly with the number of antennas ($M$).

The Time-Efficiency of Daisy Chains

The key insight of this research is the "time-optimality" of the daisy chain. Unlike the star topology, a daisy chain allows for multiple pairs of antennas to perform measurements simultaneously. For example, in a chain of antennas, pairs (1,2), (3,4), and (5,6) can calibrate in parallel. This reduces the total time required for a full sweep to a constant $4T$ seconds, regardless of the array size.

When constrained by a fixed time budget rather than a fixed number of measurements, the daisy chain can perform many more measurements than the star topology. This increased data volume compensates for the "multi-hop" error accumulation, making the daisy chain more accurate for arrays with 5 or more antennas.

Key Results and Performance Comparison

Numerical simulations for a 129-antenna base station confirm these theoretical findings. As shown in the results, when given the same time resources (allowing for more measurements in the daisy chain case, e.g., $I=64$), the daisy chain significantly outperforms the star topology.

Average MSE vs SNR for Star and Daisy Chain Interconnections

As the number of antennas $M$ grows, the average CRLB for the daisy chain topology asymptotically approaches exactly half that of the star topology under equal time constraints. This represents a substantial improvement in calibration accuracy for massive MIMO deployments.

Conclusion

This research provides a clear guideline for hardware designers of massive MIMO base stations. While star topologies offer the best accuracy per measurement, the daisy chain's ability to parallelize calibration makes it the superior choice for real-world systems operating under strict time constraints. Future work may further refine these strategies by optimizing the scheduling of parallel measurements to maximize performance within available time slots.

Summary updated

Abstract

In time-division duplexing (TDD) systems, massive multiple-input multiple-output (MIMO) relies on the channel reciprocity to obtain the downlink (DL) channel state information (CSI) with the acquired uplink (UL) CSI at the base station (BS). However, the mismatches in the radio frequency (RF) analog circuits at different antennas at the BS break the end-to-end UL and DL channel reciprocity. To restore the channel reciprocity, it is necessary to calibrate all the antennas at the BS. This paper addresses the interconnection strategy for the internal self-calibration at the BS where different antennas are interconnected via hardware transmission lines. Specifically, the paper reveals the optimality of the star interconnection and the daisy chain interconnection respectively. From the results, we see the star interconnection is the optimal interconnection strategy when the BS are given the same number of measurements. On the other hand, the daisy chain interconnection outperforms the star interconnection when the same amount of time resources are consumed. Numerical results corroborate our theoretical analyses.
BibTeX
@inproceedings{yang2018how,
  author = {Fuqian Yang and Hanyu Zhu and Cong Shen and Linglong Dai and Xiliang Luo},
  title = {{How to interconnect for Massive MIMO Self-Calibration?}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2018},
  doi = {10.1109/icassp.2018.8462296},
}
Figure